Imperfect prediction, despite being imperfect, can be valuable for decision-making purposes.
Michael Kattan, 20021(p887)
The statement by Dr Kattan1 illustrates the natural and often conflicting tensions faced by clinicians when simultaneously trying to estimate a patient’s risk, preferably in a quantitative fashion, while providing treatment, a naturally subjective behavior. The need for such tools is most apparent with the introduction of new treatments or therapeutic strategies that either challenge the status quo or preferentially benefit certain patients while exposing others to harm. The TIMI (Thrombolysis in Myocardial Infarction) risk score,2 for example, was developed and validated in studies comparing patients who received low-molecular-weight heparin with patients who received unfractionated heparin (the standard treatment). The score accurately stratified patients into having different levels of short-term risk and, as a corollary, into having variable levels of potential benefit when treated with low-molecular-weight heparin.